What is spine and leaf architecture

Spine and leaf architecture

Spine and leaf architecture are built of with combining both leaf switches and spine switches. In an architecture Every leaf switch connects to every spine switch in the network fabric. A leaf-spine topology can be layer 2 or layer 3. 

Basically, it’s useful for Data Centres that require more east-west network traffic than north-south traffic. The leaf switches connect to servers and storage devices and spine switches are connected to leaf switches. Leaf switches mesh into the spine, forming the access layer that delivers network connectivity for servers. As leaf switch connects to every spine switch in the network fabric, it minimizes the latency in the network. As well in providing many paths between two network points, all of them became able to carry traffic and thus reduces the possibility of congestion in a large network.

The key advantages of using the leaf- spine technology is as below:

• Improving Latency in the network
• Reduces congestion
• Improved scalability and flexibility

Spine Switch

A spine switch typically supports the architecture. It supports CLOS architecture. CLOS architecture is a multistage circuit switching network. For Data Centre a spine switch now a days comes with virtualization capacity which enables multiple logical device connectivity or VXLAN. They support DCBX and PFC features in all of its ports.

In layer 2 switching, it Supports LACP, STP, Jumbo Frames, IGMP, MLD, LLDP and LACP. In layer 3 routing, it supports Static routing, RIP, OSPF, GBP, IPV4, IPV6, Policy based routing, ECMP, DHCP. A leaf switch also have similar kind of protocol supports during operation.

Spine switch architecture and features:

• It Shall have dedicated management module slots in addition to the interface modules.
• It shall be designed as fully distributed architecture with separation of data and control planes to deliver enhanced   fault tolerance with zero service disruption during planned or unplanned events.
• It Shall have routing/switching capacity of minimum of 30 Tbps or more and wired speed non-blocking forwarding performance.
• Shall have the capability to extend the control plane across multiple active switches.
• making it a virtual switching fabric or equivalent feature, enabling interconnected switches to aggregate the links.
• Shall have access Control Lists for both IPv4 and IPv6 for filtering traffic to prevent unauthorized users from accessing the network.
• Shall support Port-based rate limiting and access control list (ACL) based rate limiting.
• Shall support Weighted Random Early Detection (WRED) / Random Early Detection (RED) for congestion avoidance.
• Shall have ports to support both 40G/ 100G and  100G QSFP+ equally distributed across interface modules.
• Port mirroring to duplicate port traffic (ingress and egress) to a local or remote monitoring port as a management feature needed at spine switches.
• Shall support ARP attack protection to protect against attacks that use a large number of ARP requests as well as Packet storm protection to protect against unknown broadcast unknown multicast, or unicast storms with user-defined thresholds.

-DR